技巧
在推荐数据越来越大的今天，一方面需要有好的硬件支撑，一方面是好的工具实现。2023年初这段时间，polars在赛圈越来越火就是一个例子。
模型方向
树： count feature(是否使用test的leak信息), embedding feature, target encoding
NN: 交叉特征，序列信息
召回
itemCF经典操作参考：江离cikm 2019
规则：同类别，最HOT，近期兴趣
word2vec
FAISS
YouTube-DNN
排序
优化
冷启动问题
行为序列建模
图关系
特征
降维查看user item embedding的区分度
参考与阅读
Overall
召回：https://zhuanlan.zhihu.com/p/353436475
排序：https://zhuanlan.zhihu.com/p/550513066
https://github.com/wzhe06/SparrowRecSys
https://github.com/microsoft/recommenders
https://github.com/mJackie/RecSys
https://zhuanlan.zhihu.com/p/351190043
https://github.com/xingzhexiaozhu/MovieRecommendation
https://github.com/ZiyaoGeng/Recommender-System-with-TF2.0
CTR: https://zhuanlan.zhihu.com/p/104307718
https://zhuanlan.zhihu.com/p/109933924
https://github.com/DeepGraphLearning/RecommenderSystems
https://github.com/zhaozhiyong19890102/Recommender-System
https://github.com/jihoo-kim/awesome-RecSys
https://github.com/datawhalechina/fun-rec
KDD https://github.com/aister2020?tab=repositories
https://github.com/DeepGraphLearning/RecommenderSystems
https://github.com/wangshusen/RecommenderSystem
https://github.com/ChuanyuXue/Recommender-Systems-Competition-TopSolutions
https://zhuanlan.zhihu.com/p/351190043
https://github.com/oreilly-japan/RecommenderSystems
FM: 解决数据稀疏的情况下，特征怎样组合的问题
https://www.cnblogs.com/Allen-rg/p/10750393.html
http://d2l.ai/chapter_recommender-systems/fm.html
https://www.zhihu.com/question/362190044/answer/1670206355
工程化trick：https://zhuanlan.zhihu.com/p/341452558
https://github.com/jfpuget/LibFM_in_Keras/blob/master/keras_blog.ipynb
https://zhuanlan.zhihu.com/p/58160982
https://zhuanlan.zhihu.com/p/145436595
FFM：
https://zhuanlan.zhihu.com/p/347014236
https://zhuanlan.zhihu.com/p/328481154
https://github.com/nzc/tencent-contest
SINE:
https://github.com/lambdaji/tf_repos
微信看一看： http://blog.itpub.net/31559354/viewspace-2704029/
https://www.zhihu.com/question/451498156/answer/1802577845
粗排：
https://zhuanlan.zhihu.com/p/355828527
入门：
https://tianchi.aliyun.com/competition/entrance/531842/forum
https://zhuanlan.zhihu.com/p/353436475
https://bjt.name/2018/07/03/bpr.html
https://github.com/ikaruga0508/tianchi_news_pub
https://github.com/LogicJake/tuling-video-click-top3
https://github.com/miziha-zp/BiuG-XMRec-WSDMCup22
https://github.com/opdai/wsdm2022-xmrec-top1-solution
https://github.com/gaolinjie/awesome-recommender-systems
https://github.com/tensorflow/ranking https://github.com/hongleizhang/RSPapers
https://github.com/rn5l/rsc19
https://zhuanlan.zhihu.com/p/139256086
https://zhuanlan.zhihu.com/p/353436475
https://zhuanlan.zhihu.com/p/35046241
https://github.com/ChuanyuXue/Recommender-Systems-Competition-TopSolutions
https://github.com/CharlesPikachu/algorithm/tree/master/python/SortingAlgorithm
https://neptune.ai/blog/tabular-data-binary-classification-tips-and-tricks-from-5-kaggle-competitions
Last updated